linux-hardening - Re: [PATCH] elf: Relax assumptions about vaddr ordering

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 28 Jan 2022 09:26:09 +0100
From:   Magnus Groß <magnus.gross@...h-aachen.de>
To:     Kees Cook <keescook@...omium.org>
CC:     Andrew Morton <akpm@...ux-foundation.org>,
        Alexander Viro <viro@...iv.linux.org.uk>,
        Eric Biederman <ebiederm@...ssion.com>,
        <linux-fsdevel@...r.kernel.org>, <linux-mm@...ck.org>,
        <linux-hardening@...r.kernel.org>
Subject: Re: [PATCH] elf: Relax assumptions about vaddr ordering

On Wed, Jan 26, 2022 at 10:31:42PM -0800 Kees Cook wrote:
> On Wed, Jan 26, 2022 at 08:50:15AM -0800, Kees Cook wrote:
> > On Wed, Jan 26, 2022 at 05:25:20PM +0100, Magnus Groß wrote:
> > > From ff4dde97e82727727bda711f2367c05663498b24 Mon Sep 17 00:00:00 2001
> > > From: =?UTF-8?q?Magnus=20Gro=C3=9F?= <magnus.gross@...h-aachen.de>
> > > Date: Wed, 26 Jan 2022 16:35:07 +0100
> > > Subject: [PATCH] elf: Relax assumptions about vaddr ordering
> > > MIME-Version: 1.0
> > > Content-Type: text/plain; charset=UTF-8
> > > Content-Transfer-Encoding: 8bit
> > > 
> > > Commit 5f501d555653 ("binfmt_elf: reintroduce using
> > > MAP_FIXED_NOREPLACE") introduced a regression, where the kernel now
> > > assumes that PT_LOAD segments are ordered by vaddr in load_elf_binary().
> > > 
> > > Specifically consider an ELF binary with the following PT_LOAD segments:
> > > 
> > > Type  Offset   VirtAddr   PhysAddr   FileSiz  MemSiz    Flg Align
> > > LOAD  0x000000 0x08000000 0x08000000 0x474585 0x474585  R E 0x1000
> > > LOAD  0x475000 0x08475000 0x08475000 0x090a4  0xc6c10   RW  0x1000
> > > LOAD  0x47f000 0x00010000 0x00010000 0x00000  0x7ff0000     0x1000
> > > 
> > > Note how the last segment is actually the first segment and vice versa.
> > > 
> > > Since total_mapping_size() only computes the difference between the
> > > first and the last segment in the order that they appear, it will return
> > > a size of 0 in this case, thus causing load_elf_binary() to fail, which
> > > did not happen before that change.
> > > 
> > > Strictly speaking total_mapping_size() made that assumption already
> > > before that patch, but the issue did not appear because the old
> > > load_addr_set guards never allowed this call to total_mapping_size().
> > > 
> > > Instead of fixing this by reverting to the old load_addr_set logic, we
> > > fix this by comparing the correct first and last segments in
> > > total_mapping_size().
> > 
> > Ah, nice. Yeah, this is good.
> > 
> > > Signed-off-by: Magnus Groß <magnus.gross@...h-aachen.de>
> > 
> > Fixes: 5f501d555653 ("binfmt_elf: reintroduce using MAP_FIXED_NOREPLACE")
> > Cc: stable@...r.kernel.org
> > Acked-by: Kees Cook <keescook@...omium.org>
> 
> Andrew, can you pick this up too?
> 
> -Kees
> 

May I also propose to include this patch in whatever mailing-list
corresponds to the 5.16.x bugfix series?
It turns out that almost all native Linux games published by the Virtual
Programming company have this kind of weird PT_LOAD ordering including
the famous Bioshock Infinite, so right now those games are all
completely broken since Linux 5.16.

P.S.: Someone should probably ask Virtual Programming, what kind of
tooling they use to create such convoluted ELF binaries.

> > 
> > -Kees
> > 
> > > ---
> > >  fs/binfmt_elf.c | 18 ++++++++++++++----
> > >  1 file changed, 14 insertions(+), 4 deletions(-)
> > > 
> > > diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c
> > > index f8c7f26f1fbb..0caaad9eddd1 100644
> > > --- a/fs/binfmt_elf.c
> > > +++ b/fs/binfmt_elf.c
> > > @@ -402,19 +402,29 @@ static unsigned long elf_map(struct file *filep, unsigned long addr,
> > >  static unsigned long total_mapping_size(const struct elf_phdr *cmds, int nr)
> > >  {
> > >  	int i, first_idx = -1, last_idx = -1;
> > > +	unsigned long min_vaddr = ULONG_MAX, max_vaddr = 0;
> > >  
> > >  	for (i = 0; i < nr; i++) {
> > >  		if (cmds[i].p_type == PT_LOAD) {
> > > -			last_idx = i;
> > > -			if (first_idx == -1)
> > > +			/*
> > > +			 * The PT_LOAD segments are not necessarily ordered
> > > +			 * by vaddr. Make sure that we get the segment with
> > > +			 * minimum vaddr (maximum vaddr respectively)
> > > +			 */
> > > +			if (cmds[i].p_vaddr <= min_vaddr) {
> > >  				first_idx = i;
> > > +				min_vaddr = cmds[i].p_vaddr;
> > > +			}
> > > +			if (cmds[i].p_vaddr >= max_vaddr) {
> > > +				last_idx = i;
> > > +				max_vaddr = cmds[i].p_vaddr;
> > > +			}
> > >  		}
> > >  	}
> > >  	if (first_idx == -1)
> > >  		return 0;
> > >  
> > > -	return cmds[last_idx].p_vaddr + cmds[last_idx].p_memsz -
> > > -				ELF_PAGESTART(cmds[first_idx].p_vaddr);
> > > +	return max_vaddr + cmds[last_idx].p_memsz - ELF_PAGESTART(min_vaddr);
> > >  }
> > >  
> > >  static int elf_read(struct file *file, void *buf, size_t len, loff_t pos)
> > > -- 
> > > 2.34.1
> > 
> > -- 
> > Kees Cook
> 
> -- 
> Kees Cook